Potential deadlock in SpitCase impl Stream

[clux]

interesting new clippy lint seems to suggest that impl Stream for SplitCase can deadlock (rust issue: rust-lang/rust#93883 ).

this is used in applier.

haven't had time to look at it fully, but it looks similar to our setup, and it blocks ci atm.

clippied code shows this.inner being locked
https://github.com/kube-rs/kube-rs/blob/98cf1c89c37bb8bfc12b0d31d0b3264b13076771/kube-runtime/src/utils/mod.rs#L78-L97

[clux]

actually, it looks like we are doing the right thing here? we are locking in the base of that fn. this is invalid, right?
maybe clippy is getting confused by all the shadowing of inner?

[nightkr]

Whoa, that's a thoroughly unhelpful error message.. :/

I don't think we have an issue here, since the temporaries involved here are:

1. inner.as_mut(): just a "dumb" projection of inner, misuse is protected against since it grabs a mut borrow on inner for its lifetime
2. poll_next(): stateful and complex, but the return value is "just" the ADT Poll<Option<S::Item>>

Poll and Option are both just dumb data and I don't see any has_significant_drop attribute on Pin, so I can only assume that it assumes the worst for S::Item?